Signal vs Noise

Interceptions Get the Blame. Fumbles Do the Damage.

Play for play, a lost fumble costs more win probability than an interception. We expected the opposite going in, and the data reversed us in every season since 1999.

Fan court treats the two turnovers very differently. The interception is the cardinal sin, proof of a quarterback's bad judgment. The fumble is bad luck, a freak event that could happen to anyone. We went into this expecting the numbers to back the hierarchy: an interception happens downfield and often comes with return yards, so it should cost more. The measurement came back the other way, in every single season.

More than half of lost fumbles swing the game by at least ten points of win probability. Interceptions clear that bar 43 percent of the time. The average fumble hurts more too.

The reason is about expectations, not yards. The cost of a turnover is the gap between the possession you expected to have and the one you got. Interceptions cluster on third-and-long and on deep shots, moments when the offense's expected value was already low and a punt was coming anyway; part of the loss was priced in. Lost fumbles strike on ordinary runs, completed passes, and strip sacks, in the middle of drives that were very much alive.

Many interceptions take away a possession the offense had already half given up on. A fumble takes away one it fully expected to keep.

The coin flip nobody owns

The fumble has a second property that should change how you read a season: once the ball is on the ground, nobody controls what happens next. Offenses keep about 47.5 percent of their own fumbles that draw a credited recovery, and a team's recovery share in one season carries over to the next at a correlation of about +0.01, which is to say not at all. Each team gets roughly 20 to 30 of these flips a year. If your team kept every loose ball this season, thank the bounces, and do not expect a repeat.

The part that really is a skill

Not everything in the fumble pipeline is chance, and the skill lives with individual players, not teams. Defenders who force fumbles keep forcing them: a player's forced-fumble count has repeated year to year at a correlation of about +0.26, positive in every season-to-season comparison since 1999, while team-level totals barely repeat at all. Punch-out technique travels with the player. What happens after the punch belongs to the lottery, so project the trait, not the total: an eight-strip season still regresses toward two to four.

Put it together and turnover talk gets much sharper. When a team finishes plus-12 in takeaways, split the ledger: the strips and the pressure that created the chances are real and partly repeatable, while the recovery bounces and the opponent's generosity are weather. And the next time your quarterback throws a third-and-14 interception 40 yards downfield, save some of the outrage for Sunday's real killer: the handoff that ended up on the turf at midfield.